spoonsso / dannce

MIT License
214 stars 30 forks source link

frames_with_good_tracking #149

Open yuan0821 opened 1 year ago

yuan0821 commented 1 year ago

I have 28min with 60 fps videos, total 100800 frames for each video. After running preprocess_dannce.mat, it became 504000 frames; and the move frames were 356607; Then after every 50 frames for t-sne embedding, there were only 7133 frames saved as frames_with_good_tracking.

This confused me a lot.

  1. Why do we only use 7133 frames, is it mean that we lost 95% frame's information?
  2. And how to compare the behavior map between different mice when the frames_with_good_tracking are different?
  3. we have generated the data before doing -SNE; now the issue is we cannot save all mice data together and perform post-analysis cause it is too large. You have separate 30-minute sessions, I also perform 30 min videos separately, but now we want to use all five h videos (10 x 30min) to do t-SNE; the current version of the script can't help to achieve it. Please kindly suggest what to do next.
  4. I wonder if there is any script for statistical analysis after performing CAPTURE_demo.mat. Thank you very much. @spoonsso @jessedmarshall @ @davidhildebrand